Message from Myrna Gopnik

[Frederick Newmeyer]

Dear Funknet subscribers,

Myrna Gopnik has read the exchange in Funknet regarding the 'K family' and
the general question of genetic dysphasia, and was kind enough to ask me
to forward to you the following message:

--fritz

---------- Forwarded message ----------
Date: Mon, 27 Jan 1997 12:35:42 EST5EDT
From: GOPNIK at LANGS.Lan.McGill.CA
To: fjn at u.washington.edu
Subject: reply


        Over the last few years SLI has become a hot topic because it may have
the potential to tell us something about the biological basis of
language.  In a recent exchange Liz Bates has raised some questions
about this research.  I am sorry to say that comments about this
research have often generated more heat than light. If we are really
interested in the science of it all then it is important to get the
issues out on the table and see which ones we can agree about, which
ones are still outstanding and how we could resolve them.  So this note
is intended not so much to cite data as it is to at least make a stab at
clarifying some of the issues. (I will not be able to resist citing a
little data and I would be glad to respond to any request for more
details.)  Our research program over the years has been clear:  start
with broad ranging, linguistically significant tests; examine the
results; construct linguistically sound hypotheses; design new, hard
tests, which sometimes require looking at new languages; look at the
results; refine the hypotheses and start all over again.  And it has
worked.  Bates appears to fault us for not always using standardized
tests, but those tests are useless for addressing new hypotheses.  For
example, the original data from English, Japanese and Greek told us that
language impaired subjects had particular trouble with inflections like
tense.  The linguistic question was whether it was inflectional rules or
morphological complexity that was the problem.  Would they have as much
difficulty in finding the root in a complex word as they clearly had in
adding an inflection?  There was no way to test this hypothesis in
English, but Jenny Dalalakis pointed out that there was a way of testing
it in Greek; you have to be able to extract roots out of inflected forms
in order to construct new compounds or diminutives.  No one had ever
looked at this before so she had to construct totally new tests, find
out if and when young children could do these tasks, and then try them
out on impaired unilingual native speakers of Greek.  Jenny Dalalakis's
innovative work on compounds and diminutives in Greek has made it clear
that the language impaired subjects have just as much difficulty with
finding the root of a complex word as they have with adding an
inflection.  And they have these problems in nouns and adjectives and
not just on verbs so the problem cannot be accounted for by agreement or
optional infinitives.  Had we taken the easy route and just stuck with
standardized tests we would not have been able to address these
linguistically significant questions.  Testing new hypotheses, with new
tests on new languages is a risky business, but it is the necessary path
if you want to find out new truths and not merely confirm old models.
The aim, after all, is not merely to tally up the numbers from tests,
but to use these results to construct hypotheses about the internal
grammar of the individual that is producing these results.  From this
point of view the contrast of "deviant" vs. "delay" is not easily
interpretable.  We know  that at an early stage of language development
children treat inflected words as if they were unanalyzed chunks and it
looks like language impaired individuals do the same.  But there are
other huge differences in their grammars with respect to the lexicon,
syntax and compensatory strategies.   On what grounds can this one
particular similarity lead us to say that their grammar is "delayed" and
not "deviant" (especially since we know that this "delay" lasts at least
until age 76)?
          After eight years,  hundreds of tests, thousands of data points, and
almost a hundred impaired subjects representing four different native
languages (English, French, Greek and Japanese), we are convinced that
the data converges to tell us that, among other things, the language
impaired subjects cannot construct normal representations for
grammatically complex words and they therefore cannot use rules which
depend on the content of these representations.  Let me give you just
one example of the kind of cross-linguistic evidence that we have
gathered.  If a subject produces a form like "walked" that appears to be
inflected, we cannot tell whether this form is grammatically complex in
their mental lexicon or whether it is merely a single unanalyzed chunk.
One of the ways we have studied this is to see if, given a novel word,
the subject could produce complex novel forms.

Ability to mark novel words grammatically

In each of these tests the subjects were given a context, usually in
pictures,  which required that a grammatical rule be applied to a novel
word: This pencil is weff.  This pencil is even  _____.


        % correct                   CONTROLS         IMPAIRED
        PAST TENSE
        English (in England)         95.4               38.0
        English (in Canada)       93.5          52.3
        Greek                       87.1                20.0
        French                     92.6         33.3
        Japanese                        89.1            37.0
        PLURALS
        English (in England)         95.7               57.0
        English (in Canada)           99.2              58.3
        Greek                       79.8                42.1
 COMPARATIVES
        English (in England)         74                 21
        COMPOUNDS
        Japanese                         80.5           20.2
        Greek                           93.6            12.8
        DIMINUTIVES
        Greek                       83.9                40.2


        These data clearly and convincingly show that the language impaired
subjects are significantly worse at producing complex forms from novel
words than are unimpaired subjects for every grammatical task and for
every language population that we have tested. (We have closets of data
and drawers of papers about a wide range tests that we would be glad to
send to anyone who is interested.  N.B. we have already sent out over
200 copies of the McGill working papers that Bates refers to, so it is
not exactly hard to get).    Are we finished with our work?  Not by a
long shot.  We have lots more questions still to be answered just about
morphology and though we know that they have problems with  prosody and
syntax we do not understand the complete picture yet.  There are more
things in language impairment than are accounted for in our hypotheses.
But we think that our program of research has made real progress and
that eventually the truth will out.
          Bates once more raises two specific issues with respect to the K
family: oral apraxia and low IQ.  I have responded to them often before,
but they are harder to put to rest than old Hamlet's ghost.   I will try
once more.  There seems to be three different ways in which these issues
have been used 1. to imply that I have misreported the facts about the
K. family and therefore my work should not be trusted, 2.  to claim that
the apraxia (Fletcher, 1990) or the low IQ (Vargha-Khadem) are the
direct cause of the language problems, 3. to argue that this is not
primarily a disorder of language because the members of the K family
have other problems.  The first issue is easy to clarify.  I am a
linguist, therefore I depended on reports from other experts with
respect to apraxia and IQ.  A neurologist, Dr. T. Ashizawa, examined the
oral movements of several of the language impaired members of the K.
family in the standard way and reported that they had normal oral motor
function.   Vargha-Khadem herself reports that the signs of apraxia show
up only when the subjects have to perform several oral activities at the
same time.   The original IQ tests which I reported were done by the
school psychologists who, over several years, found that all of the
children were in the normal range.  Vargha-Khadem later retested them
with a much wider range of newly normed tests and the numbers changed.
Do they have apraxia? Do they have low performance IQ?  It depends on
what tests you use and what your criteria are.  All we can conclude is
that if you use different tests at different times with different
criteria you can get different results -- there needs no Bates, my
friends, come from California to tell us this.  But what does it all
mean?  First apraxia.  Two phonologists, G. Piggott and H. Goad, spent
many hours with the K family and even more hours listening to and
transcribing tapes.  They conclude that the errors that they make when
they talk cannot be accounted for by oral motor problems (this is all
reported on in detail in the McGill Working Papers).  In fact, these
subjects regularly produce forms that are harder to articulate than the
normal form.  One example: in producing the plural forms for novel words
they regularly do not use voicing assimilation, i.e. they say /wug-s/
not /wug-z/.  Fletcher early on suggested that their problems with tense
might be accounted for by the articulatory vulnerability of regular
tense marking in English.  We tested this hypothesis and it is simply
false. We have reported on hundreds of pieces of data about their use of
tense in a wide range of different tests; spontaneous speech, reading,
writing and they show the same pattern of problems with tense across all
of these tests.  Bates, herself, now seems to grant that the linguistic
data cannot be accounted for by any purported oral motor problems.  So
even if one grants that some members of the K. family may have oral
apraxia it tells us nothing about their language problems.  Let's get
beyond the K family for a minute.  This problem with tense is not unique
to them.  It shows up everywhere that we have looked:

Ability to produce tense marking

% correct
               English (Eng.)  English (Can.)    French         Japanese          Greek
impaired          38.3                  52.3            46.7            48.1        20.0
controls          91.7              93.5                96.4        97.9            87.1

Subjects were given items like:  Everyday I walk to school.  Just like
everyday, yesterday I ______ . This task requires the subject to
recognize that the temporal context specified in the second sentence
requires a particular verb form.

Language impaired subjects outside of the K family  have similar
problems with tense  no matter how it is encoded in their native
language: in a final stressed syllable as in French or in a three
syllable element like "mashita" in Japanese

        The IQ scores  tell us nothing about their problems with language
either.  Even if we credit the new K. numbers over the original numbers
all they show is that the means for the two groups are different.  But
still a language impaired member of the K. family scores 112, while his
unimpaired relative scores only 84.  It is a truism in speech pathology
that some individuals with language disorder have lowish performance IQs
and others have very high scores, as high as 140.  (I have even recently
been told about a language impaired subject who is a rocket scientist.)
And there are lots of cases of individuals with low performance IQ who
do not have these kinds of problems with language.  Performance IQ
scores simply do not correlate with these patterns of language
impairment.
        It is absolutely clear, from our own research as well as from other
research,  that  individuals with language impairment sometimes have
other problems ranging from dyslexia to spatial rotation to depression
to schizophrenia to apraxias and agnosias, but none of these other
specific deficits reliably occurs with the language disorder and there
are many individuals that have one of these other problems without
having any language disorder.  Language impaired let us grant them then,
and now remains that we find out the cause of this effect, or rather say
the cause of this defect.  The question is whether the language
impairment that we see in these individuals comes from a separate and
special "language faculty" that is out of order or whether some more
general cognitive or perceptual processing system is not functioning and
the purported "language" problems are merely a result of a breakdown in
a much more general system.
                Do these other problems simply co-occur with this disorder or are
they the proximate cause of the language problems?  What mechanisms could
make disorders co-occur? One possibility is the contiguity of the genes
that code for the co-occurring disorders.  For example, the most striking
feature of individuals with William's syndrome is that their spatial
reasoning is seriously impaired.  It is also the case that many of these
individuals have heart and joint problems (as do many individuals who do
not have spatial problems).  But no one seriously supposes that their
heart and joint problems cause their spatial problems.  What appears to be
going on is that the gene that is implicated in William's syndrome is very
close to the genes that code for elastin which build normal hearts and
joints.  If the mutation is large it hits several genes at the same time.
        Though there is now a flood of epidemiological studies and twin
studies that all indicate that at least some cases of SLI are connected
to some genetic factors, it is still a puzzle about what the exact
pattern of inheritance is and what genes are involved--lots of groups
are working on these problems including our team lead by Roberta
Palmour.  There will be a loud hurrah! when this is figured out.  But
since we do not yet know what the genes are we cannot tell if the
co-occurring problems are accountable for by genetic contiguity.
        The natural next question is what direct effect do these genes have on
the organism?  The obvious answer seems to be that something goes wrong
in the development of the brain.  Elena Plante and her colleagues have
found significant differences in brain structure associated with this
disorder.  Alan Evans, Terry Peters and Noor Kabani,  part of our team
at the Montreal Neurological Institute,  are studying MRI data from our
Canadian subjects and, though the analysis is still ongoing, preliminary
data shows that our impaired subjects also have significant
neuroanatomical anomalies (in press, Journal of Neurolinguistics).
Tanya Gallagher and Ken Watkin, also of McGill, have just completed an
ultra-sound study comparing brain development in a fetus with a positive
family history of SLI to three fetuses with no family history of
language impairment.  These studies show that there are significant
differences between the two groups in the pattern of brain development
in the last trimester of pregnancy and that these differences involve
those centers of the brain that have been implicated for language (in
press,  Journal of Neurolinguistics).  So at least for now it looks like
there are reasons to believe that at least some cases of developmental
language impairment are inherited and involve neurological anomalies.
It seems likely that a possible source of the wide variety of other
problems that we see in some particular individuals with language
impairment is that the neurological damage caused by this mutation may
affect different parts of the brain in different individuals.   But this
is just a hunch.  Lots more about the nature of the genetic and
neurological patterns that are associated with this language impairment
still has to be understood.   We are working on it.  So are lots of
others.

        The question that concerns me as a linguist is what precisely and
exactly is wrong with their language and, more importantly, is something
wrong with their ability to acquire language itself or are their
language problems just an epiphenomenon of some other, non-linguistic
problem.  The only way to answer this question is to look at what they
do and how they do it and why they do it and then come up with an
explanatory theory that accounts for these facts.  That is what we have
been doing and the data converge to tell us that the language faculty is
directly affected in these subjects.